Quantification of blood loss using the Hb/kg index in cardiac surgery

Background. Perioperative blood loss in cardiac surgery remains a significant clinical challenge as there is no “gold standard” for its accurate assessment [1, 2]. Conventional volume-based methods often underestimate or overestimate actual blood loss [3], especially given individual differences in body mass and chest tubes composition (haemorrhagic vs. serous) [4]. We hypothesized that blood loss could be more accurately calculated using the Hb/kg index in terms of haemoglobin (Hb) mass loss per kilogram the patient's body weight. Objective: To develop a novel approach for calculating actual blood loss using the Hb/kg index. Methods. We studied 195 patients undergoing cardiac surgery over 13 months (October 2023–November 2024). The Hb/kg index was calculated using intraoperative Hb loss, Hb loss via chest tubes, packed red blood cell transfusions, and patient weight. Eighty-six additional clinical predictors were analyzed using conventional statistics and machine learning algorithms, including Linear Regression, Lasso, Lasso-OLS, Support Vector Machine, k-Nearest Neighbors, Decision Tree, XGBoost, and Deep Neural Network. Predictors with statistically significant Spearman correlations with the Hb/kg index were included for further analysis. Results: Lasso regression achieved the best overall performance in predicting Hb/kg index. It yielded the lowest mean squared error (0.08 ± 0.04), and mean absolute percentage error (0.18 ± 0.10), with the highest correlation (0.92 ± 0.06) and R² score (0.82 ± 0.13). BMI showed a significant negative relationship (-0.018, p < 0.001). Postoperative Hb and haematocrit values had negative correlation (-0.69, p<0.001 and -0.07, p<0.015), while initial Hb was positively correlated (0.85, p<0.001). Conclusion: This method provides a more reliable and clinically relevant tool to calculate actual blood loss and allows for a more precise assessment and treatment. This approach could be a robust research tool, although further studies are needed to demonstrate the way serves as a predictor of surgical and clinical outcomes.